Smoothing with Fake Label

Ziyang Luo; Yadong Xi; Xiaoxi Mao

doi:10.1145/3459637.3482184

Smoothing with Fake Label

Ziyang Luo, Yadong Xi, Xiaoxi Mao

Department of Computer Science

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

1 Citation (Scopus)

Abstract

Label Smoothing is a widely used technique in many areas. It can prevent the network from being over-confident. However, it hypotheses that the prior distribution of all classes is uniform. Here, we decide to abandon this hypothesis and propose a new smoothing method, called Smoothing with Fake Label. It shares a part of the prediction probability to a new fake class. Our experiment results show that the method can increase the performance of the models on most tasks and outperform the Label Smoothing on text classification and cross-lingual transfer tasks.

Original language	English
Title of host publication	CIKM 2021 - Proceedings of the 30th ACM International Conference on Information and Knowledge Management
Publisher	Association for Computing Machinery (ACM)
Pages	3303-3307
Number of pages	5
ISBN (Electronic)	9781450384469
DOIs	https://doi.org/10.1145/3459637.3482184
Publication status	Published - 30 Oct 2021
Event	30th ACM International Conference on Information and Knowledge Management, CIKM 2021 - Virtual, Online, Gold Coast, Queensland, Australia Duration: 1 Nov 2021 → 5 Nov 2021 https://www.cikm2021.org/ https://dl.acm.org/doi/proceedings/10.1145/3459637

Publication series

Name	Proceedings of the International Conference on Information and Knowledge Management
Publisher	Association for Computing Machinery

Conference

Conference	30th ACM International Conference on Information and Knowledge Management, CIKM 2021
Country/Territory	Australia
City	Gold Coast, Queensland
Period	1/11/21 → 5/11/21
Internet address	https://www.cikm2021.org/ https://dl.acm.org/doi/proceedings/10.1145/3459637

User-Defined Keywords

cross-lingual
label smoothing
machine translation
neural networks
text classification

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1145/3459637.3482184

Cite this

@inproceedings{88fd5e8ad18d4faba174035d61373659,

title = "Smoothing with Fake Label",

abstract = "Label Smoothing is a widely used technique in many areas. It can prevent the network from being over-confident. However, it hypotheses that the prior distribution of all classes is uniform. Here, we decide to abandon this hypothesis and propose a new smoothing method, called Smoothing with Fake Label. It shares a part of the prediction probability to a new fake class. Our experiment results show that the method can increase the performance of the models on most tasks and outperform the Label Smoothing on text classification and cross-lingual transfer tasks.",

keywords = "cross-lingual, label smoothing, machine translation, neural networks, text classification",

author = "Ziyang Luo and Yadong Xi and Xiaoxi Mao",

note = "Publisher Copyright: {\textcopyright} 2021 ACM.; 30th ACM International Conference on Information and Knowledge Management, CIKM 2021 ; Conference date: 01-11-2021 Through 05-11-2021",

year = "2021",

month = oct,

day = "30",

doi = "10.1145/3459637.3482184",

language = "English",

series = "Proceedings of the International Conference on Information and Knowledge Management",

publisher = "Association for Computing Machinery (ACM)",

pages = "3303--3307",

booktitle = "CIKM 2021 - Proceedings of the 30th ACM International Conference on Information and Knowledge Management",

address = "United States",

url = "https://www.cikm2021.org/, https://dl.acm.org/doi/proceedings/10.1145/3459637",

}

Luo, Z, Xi, Y & Mao, X 2021, Smoothing with Fake Label. in CIKM 2021 - Proceedings of the 30th ACM International Conference on Information and Knowledge Management. Proceedings of the International Conference on Information and Knowledge Management, Association for Computing Machinery (ACM), pp. 3303-3307, 30th ACM International Conference on Information and Knowledge Management, CIKM 2021, Gold Coast, Queensland, Australia, 1/11/21. https://doi.org/10.1145/3459637.3482184

Smoothing with Fake Label. / Luo, Ziyang; Xi, Yadong; Mao, Xiaoxi.
CIKM 2021 - Proceedings of the 30th ACM International Conference on Information and Knowledge Management. Association for Computing Machinery (ACM), 2021. p. 3303-3307 (Proceedings of the International Conference on Information and Knowledge Management).

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

TY - GEN

T1 - Smoothing with Fake Label

AU - Luo, Ziyang

AU - Xi, Yadong

AU - Mao, Xiaoxi

PY - 2021/10/30

Y1 - 2021/10/30

N2 - Label Smoothing is a widely used technique in many areas. It can prevent the network from being over-confident. However, it hypotheses that the prior distribution of all classes is uniform. Here, we decide to abandon this hypothesis and propose a new smoothing method, called Smoothing with Fake Label. It shares a part of the prediction probability to a new fake class. Our experiment results show that the method can increase the performance of the models on most tasks and outperform the Label Smoothing on text classification and cross-lingual transfer tasks.

AB - Label Smoothing is a widely used technique in many areas. It can prevent the network from being over-confident. However, it hypotheses that the prior distribution of all classes is uniform. Here, we decide to abandon this hypothesis and propose a new smoothing method, called Smoothing with Fake Label. It shares a part of the prediction probability to a new fake class. Our experiment results show that the method can increase the performance of the models on most tasks and outperform the Label Smoothing on text classification and cross-lingual transfer tasks.

KW - cross-lingual

KW - label smoothing

KW - machine translation

KW - neural networks

KW - text classification

UR - http://www.scopus.com/inward/record.url?scp=85119210540&partnerID=8YFLogxK

U2 - 10.1145/3459637.3482184

DO - 10.1145/3459637.3482184

M3 - Conference proceeding

AN - SCOPUS:85119210540

T3 - Proceedings of the International Conference on Information and Knowledge Management

SP - 3303

EP - 3307

BT - CIKM 2021 - Proceedings of the 30th ACM International Conference on Information and Knowledge Management

PB - Association for Computing Machinery (ACM)

T2 - 30th ACM International Conference on Information and Knowledge Management, CIKM 2021

Y2 - 1 November 2021 through 5 November 2021

ER -

Smoothing with Fake Label

Abstract

Publication series

Conference

User-Defined Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this